New Computer Algorithm May Drive Bible Scholarship a Long Distance
Date: 2011-06-30 Hour: 11:44
For centuries Bible scholars have attempted to identify distinct authorial voices within books of the Bible. Their methods have often been criticized as subjective and even impressionistic.
Now a team of computer scientists and Bible scholars has found a way to put such authorial analysis on a firm computational basis. The team – led by Bar-Ilan University computer scientist Moshe Koppel and including his student, Navot Akiva, Tel Aviv University computer scientist Nachum Dershowitz, and Hebrew University graduate student in Bible Idan Dershowitz – has devised a novel algorithm that automatically divides a composite document into distinct authorial strands.
The methodology is based on automatically identifying small clusters of text that are distinguishable from each other due to different choices among available synonyms. For example, one cluster will consistently use the word makel (meaning “staff”), while the other cluster will consistently use mateh (with the same meaning). By formalizing and generalizing this phenomenon, the researchers were able to show that when two Bible books, such as Jeremiah and Ezekiel, were randomly mixed together, the merged book could be automatically, and almost perfectly, separated out to its constituent components.
This result is of potentially overwhelming importance to Bible scholarship: when applied to books for which the underlying authorial components are unknown, the method can automatically identify the best possible such division.
The details of the method and the proof of its efficacy were presented for the first time earlier this month at the annual meeting of the Association for Computational Linguistics in Portland, Oregon. The researchers are currently working on a follow-up paper that will detail the implications of their work for Bible scholarship.
Several years ago Prof. Koppel and colleagues developed a computer algorithm that can examine an anonymous text and determine, with accuracy rates of better than 80 percent, whether the author is male or female.